Me pēhea te ako i te AI?

Me pēhea te ako i te AI?

Ko te ako i te AI he rite ki te takahi ki roto i tētahi whare pukapuka nui e karanga ana ngā pukapuka katoa "TĪMATA I KŌNEI." Ko te haurua o ngā whata pukapuka e kī ana "pāngarau," he… he āhua whanoke 😅

Ko te painga: kāore koe e hiahia kia mōhio ki ngā mea katoa hei hanga i ngā mea whai hua. Me whai ara whakaaro nui koe, me ētahi rauemi pono, me te hiahia kia pōhēhē mō tētahi wā poto (ko te pōhēhē te utu whakauru).

Ngā tuhinga ka pai pea koe ki te pānui i muri i tēnei:

🔗 Me pēhea te kite a te AI i ngā āhuatanga rerekē
E whakamārama ana i ngā tikanga kimi i ngā āhuatanga rerekē mā te whakamahi i te ako mīhini me ngā tatauranga.

🔗 He aha te kino o te AI mō te hapori
Ka arotake i ngā mōrearea matatika, pāpori, me te ōhanga o te mātauranga horihori.

🔗 E hia te wai e whakamahia ana e AI
Ka wetewete i te whakamahinga pūngao o te AI me ngā pānga huna o te whakamahinga wai.

🔗 He aha te huingararaunga AI
E tautuhi ana i ngā huinga raraunga, te tapanga, me tō rātou tūranga i roto i te whakangungu i te AI.

Te tikanga tūturu o te "AI" i ngā kupu o ia rā 🤷♀️

E ai ki te kōrero a te tangata mō te kupu "AI", he maha ngā tikanga:

  • Te Ako Mīhini (ML) – ka ako ngā tauira i ngā tauira mai i ngā raraunga hei mahere i ngā whakaurunga ki ngā putanga (hei tauira, te kimi pāme, te matapae utu). [1]

  • Te Ako Hōhonu (DL) – he huinga iti o te ML e whakamahi ana i ngā whatunga io i te tauine (tirohanga, kōrero, tauira reo nui). [2]

  • AI Whakaputa – ngā tauira e whakaputa ana i ngā tuhinga, ngā whakaahua, ngā waehere, ngā oro (ngā karetao kōrerorero, ngā kaiurungi tahi, ngā taputapu ihirangi). [2]

  • Te Ako Whakakaha – te ako mā te whakamātautau me te utu (ngā kaiwhakahaere kēmu, te karetao). [1]

Kāore koe e hiahia ki te whiriwhiri tika i te tīmatanga. Kaua e whakaarohia te AI he whare taonga. He rite tonu ki te kīhini - ka tere ake te ako mā te tunu kai. I ētahi wā ka tahuna e koe te tōhi. 🍞🔥

He kōrero poto: i tukuna mai e tētahi tīma iti tētahi tauira "tino pai" mō te "churn"… tae noa ki te kitenga i ngā ID ōrite i roto i te tereina me te whakamātautau. He turuturu tawhito. Nā tētahi paipa māmā me te wehenga ma i huri te 0.99 whakapae ki tētahi kaute pono (iti iho!) me tētahi tauira e tino whānui ana. [3]

He aha te mea e hanga ana i tētahi mahere pai mō te "Me pēhea te ako i te AI" ✅

He iti noa ngā āhuatanga o tētahi mahere pai e ahua hoha ana, engari e whakaora ana i a koe i ngā marama:

  • Hangaia ngā kaupapa iti i a koe e ako ana (ngā kaupapa iti i te tīmatanga, ngā kaupapa nunui i muri mai).

  • Akohia te pāngarau iti rawa e hiahiatia ana , kātahi ka porowhitahia whakamuri mō te hohonu.

  • Whakamāramahia tāu i mahi ai (tāpirihia tō mahi ki te parera rapa; ka rongoā i te whakaaro pōuri).

  • Puritia ki tētahi "pūranga matua" mō tētahi wā (Python + Jupyter + scikit-learn → kātahi ko PyTorch).

  • Inehia te ahunga whakamua mā ngā putanga , kaua mā ngā hāora e mātakihia ana.

Mena he ataata me he tuhipoka anake tāu mahere, he rite ki te ngana ki te kauhoe mā te pānui i ngā kōrero mō te wai.

Kōwhiria tō ara (mō nāianei) – e toru ngā ara noa 🚦

Ka taea e koe te ako i te AI i roto i ngā "āhua" rerekē. Anei ētahi e toru e whai hua ana:

1) Te ara hanga mahi 🛠️

Pai rawa atu mēnā e hiahia ana koe ki ngā wikitoria tere me te hihiri.
Arotahi: ngā huinga raraunga, ngā tauira whakangungu, ngā whakaaturanga tuku.
Ngā rauemi tīmatanga: Akoranga Hurihuri ML a Google, Kaggle Learn, fast.ai (ngā hononga kei roto i ngā Tohutoro me ngā Rauemi i raro nei).

2) Ko te huarahi taketake te tuatahi 📚

Pai rawa atu mēnā he pai ki a koe te mārama me te ariā.
Arotahi: te whakatauira, te whakahāngaitanga-rerekētanga, te whakaaro tūponotanga, te arotautanga.
Ngā Punga: Ngā rauemi a Stanford CS229, MIT Intro to Deep Learning. [1][2]

3) Te ara kaiwhakawhanake taupānga gen-AI ✨

He pai rawa atu mēnā kei te hiahia koe ki te hanga kaiawhina, ki te rapu, ki ngā rerengamahi, ki ngā mea "kaihoko-ā-ringa".
Arotahi: te akiaki, te kimi, ngā arotake, te whakamahinga taputapu, ngā kaupapa matua mō te haumaru, te whakatinanatanga.
Ngā tuhinga hei pupuri tata: ngā tuhinga tūāpapa (API), te akoranga HF (te whakamahi taputapu).

Ka taea e koe te whakawhiti ara ā muri ake nei. Ko te tīmatanga te wāhanga uaua.

 

Me pēhea te ako i te AI

Ripanga Whakataurite – ngā huarahi pai ki te ako (me ngā āhuatanga rerekē pono) 📋

Utauta / Akoranga hunga whakarongo Utu Te take e whai hua ai (he kōrero poto)
Akoranga Whakaterenga Ako Mīhini a Google ngā tīmatanga Koreutu Tirohanga + mahi ā-ringa; ka karo i ngā uauatanga nui
Kaggle Learn (Whakataki + Waenga ML) ngā tīmatanga e pai ana ki te whakaharatau Koreutu Ngā akoranga iti + ngā mahi whakakori tinana inamata
Te Ako Hōhonu Whaihua a fast.ai ngā kaihanga me ētahi waehere Koreutu Ka whakangungu wawe koe i ngā tauira tūturu - pērā tonu 😅
Te Ako Hōhonu. Te Tohungatanga AI ML ngā ākonga hanganga Utu Te ahunga whakamua mārama i roto i ngā ariā matua o te ML
Ngā Whakatakotoranga Ako Hōhonu.AI Ngā kaupapa matua o te ML kua oti kē Utu Te hohonutanga pakari o ngā kupenga io + ngā rerengamahi
Ngā tuhipoka a Stanford CS229 e ārahihia ana e te ariā Koreutu Ngā kaupapa matua nui ("he aha tēnei e mahi ai")
Aratohu Kaiwhakamahi scikit-learn Ngā kaimahi ML Koreutu Ko te kete taputapu matarohia mō ngā ripanga/ngā rārangi turanga
Ngā Akoranga PyTorch ngā kaihanga ako hōhonu Koreutu Ara horoi mai i ngā tensor → ngā koropiko whakangungu [4]
Akoranga LLM mō te Kanohi Awhi Ngā kaihanga NLP + LLM Koreutu Te reremahi mahi LLM whaihua + ngā taputapu rauropi
NIST AI Anga Whakahaere Morearea tetahi e whakamahi ana i te AI Koreutu He tūāpapa whakahaere mōrearea/tūponotanga māmā, whai hua hoki [5]

He kōrero poto: he rerekē te "utu" i runga ipurangi. Ko ētahi mea he kore utu engari he utu nui te aro atu… he kino ake tēnei i ētahi wā.

Ngā pūkenga matua e hiahiatia ana e koe (me te raupapatanga) 🧩

Mena ko tāu whāinga ko te ako i te AI me te kore e toremi, whaia tēnei raupapa:

  1. Ngā kaupapa matua o Python

  • Ngā mahi, ngā rārangi/ngā kupu, ngā akomanga māmā, ngā kōnae pānui.

  • Te tikanga me whai: tuhia he tuhinga poto, kaua ko ngā pukatuhi anake.

  1. Te whakahaere raraunga

  • He whakaaro rite ki a NumPy, ngā kaupapa matua o te panda, te whakatakoto mahere.

  • Ka roa te wā e noho ana koe i konei. Ehara i te mea ataahua, engari koinei te mahi.

  1. ML Matarohia (te mana nui kāore i te tino manakohia)

  • Ngā wehenga tereina/whakamātautau, te turuturu, te taunga rawa.

  • Te whakatauira rārangi/whakahaere, ngā rākau, ngā ngahere matapōkere, te whakanui i te rōnaki.

  • Ngā Inenga: te tika, te tino tika/te maumahara, ROC-AUC, MAE/RMSE - kia mōhio ai koe ki te wā e tika ana ia mea. [3]

  1. Te ako hōhonu

  • Ngā pūngao, ngā pikinga/te tautoko tuara (ā-kaupapa), ngā koropiko whakangungu.

  • Ngā CNN mō ngā whakaahua, ngā whakawhiti mō ngā tuhinga (ā tōna wā).

  • He nui te āwhina o ētahi o ngā kaupapa matua o PyTorch. [4]

  1. Ngā rerengamahi AI Whakaputa + LLM

  • Te whakawhiwhinga tohu, te whakauru, te whakaputanga whakarei ake i te tiki mai, te aromatawai.

  • Te whakatikatika pai, te akiaki rānei (ā, ina kore rawa e hiahiatia ana tētahi o aua tohutohu).

He mahere taahiraa-i-te-taahiraa ka taea e koe te whai 🗺️

Wāhanga A – kia tere te mahi a tō tauira tuatahi ⚡

Whāinga: whakangungu i tētahi mea, inehia, whakapaitia.

  • Whakatakotoria he whakataki poto (hei tauira, ML Crash Course), kātahi ka mahi i tētahi akoranga poto (hei tauira, Kaggle Intro).

  • Whakaaro kaupapa: matapae i ngā utu whare, te nekenga o ngā kiritaki, te mōrearea nama rānei i runga i tētahi huinga raraunga tūmatanui.

He rārangi arowhai iti mō te "wikitoria":

  • Ka taea e koe te uta raraunga.

  • Ka taea e koe te whakangungu i tētahi tauira taketake.

  • Ka taea e koe te whakamārama i te overfitting i te reo māmā.

Wāhanga B – kia taunga ki te mahi ML tūturu 🔧

Whāinga: kati te ohorere i ngā momo rahunga e kitea whānuitia ana.

  • Mahihia ngā kaupapa ML waenga: ngā uara ngaro, te turuturu, ngā paipa, CV.

  • Pānuihia ētahi wāhanga o te Aratohu Kaiwhakamahi scikit-learn, kātahi ka whakahaere i ngā wāhanga poto. [3]

  • Whakaaro kaupapa: he pūnaha whakahaere māmā mai i te tīmatanga ki te mutunga me te tauira kua tiakina me te pūrongo aromatawai.

Wāhanga C – te ako hōhonu kāore e rite ki te mahi makutu 🧙♂️

Whāinga: whakangungu i tētahi kupenga io me te mārama ki te takahuri whakangungu.

  • Mahia te ara "Akohia ngā Kaupapa Taketake" a PyTorch (tensors → datasets/dataloaders → training/eval → saving). [4]

  • Mena kei te hiahia koe ki te tere me te kare ā-tinana, ka taea te hono atu ki a fast.ai.

  • Whakaaro kaupapa: he whakarōpū whakaahua, he tauira kare ā-roto, he whakatikatika rānei i tētahi whakawhiti iti.

Wāhanga D – ngā taupānga AI whakaputa e mahi pono ana ✨

Whāinga: hanga i tētahi mea e whakamahia ana e te tangata.

  • Whāia he akoranga LLM mahi-ā-ringa + he tīmatanga tere mō te kaihoko ki te hono i ngā tāpiringa, te tiki mai, me ngā whakatupuranga haumaru.

  • Whakaaro kaupapa: he karetao pātai me ngā whakautu i runga i ō tuhinga (wāhanga → tāmau → tiki → whakautu me ngā tohutoro), he kaiawhina tautoko kiritaki rānei me ngā waeatanga taputapu.

Te wāhanga "pāngarau" – akohia pērā i te whakakī kinaki, kaua te kai katoa 🧂

He mea nui te pāngarau, engari ko te wā te mea nui ake.

Te iti rawa o te pāngarau ka taea hei tīmatanga:

  • Te arorangi rārangi: ngā whārite, ngā matihiko, ngā hua ira (te māramatanga mō ngā tāpiringa). [2]

  • Tātaitai: te māramatanga taupatupatu (ngā pikinga → ngā pikinga). [1]

  • Te tūponotanga: ngā tohatoha, ngā tumanakohanga, ngā whakaaro taketake o Bayes. [1]

Ki te hiahia koe ki tētahi kaupapa matua whaimana ake i muri mai, tirohia ngā tuhipoka CS229 mō ngā kaupapa matua me te ako hohonu whakataki a MIT mō ngā kaupapa hou. [1][2]

Ngā kaupapa hei whakaatu i a koe e mōhio ana ki tāu e mahi nei 😄

Ki te hanga noa koe i ngā whakarōpūtanga raraunga taonga takaro, ka noho pūmau koe. Whakamātauria ngā kaupapa e rite ana ki te mahi tūturu:

  • Kaupapa ML tūāpapa-tuatahi (scikit-learn): raraunga ma → tūāpapa pakari → tātari hapa. [3]

  • LLM + taupānga tiki: ingest docs → chunk → embed → retrieve → generate responses with quotes.

  • Papatohu aroturuki tauira iti: takiuru i ngā tāurunga/putanga; aroturuki i ngā tohu nekeneke (ā, he āwhina anō hoki ngā tatauranga māmā).

  • Arotake poto AI kawenga: tuhi i ngā mōrearea, ngā take taha, ngā pānga rahunga; whakamahia he anga māmā. [5]

Te whakatinanatanga kawenga me te mahi (āe, ahakoa mō ngā kaihanga takitahi) 🧯

He whakamātautau pono: he māmā ngā whakaaturanga whakamiharo; kāore ngā pūnaha pono.

  • Puritia he README poto, he momo "kāri tauira": ngā pūtake raraunga, ngā inenga, ngā herenga e mōhiotia ana, te whakahou i te tere.

  • Tāpirihia ngā ārai taketake (ngā herenga reiti, te whakamana tāuru, te aroturuki i te tukino).

  • Mō ngā mea e pā ana ki te kaiwhakamahi, e whai pānga ana rānei, whakamahia he e hangai ana ki te mōrearea : tautuhia ngā kino, whakamātautauhia ngā take taha, me te tuhi i ngā whakaiti. Kua hangaia te NIST AI RMF mō tēnei. [5]

Ngā mahanga noa (kia taea ai e koe te karo i aua mahanga) 🧨

  • Te pekepeke akoranga – ka huri te "kotahi atu akoranga" hei tuakiri mōu.

  • Ka tīmata ki te kaupapa tino uaua - he rawe ngā "transformers", engari ko ngā mea taketake te utu reti.

  • Te kore aro ki te aromatawai – ko te tika anake ka taea te kī he tika te mahi. Whakamahia te ine tika mō te mahi. [3]

  • Kaua e tuhia ngā mea – tuhia ngā kōrero poto: ngā mea i rahua, ngā mea i rerekē, ngā mea i pai ake.

  • Kāore he mahi whakatinanatanga – he nui ngā mea ka akohia e te takai taupānga māmā noa iho.

  • Te karo i te whakaaro mōrearea – tuhia kia rua ngā tohutohu mō ngā kino pea i mua i te tuku. [5]

Ngā Kōrero Whakamutunga – He roa rawa, kāore au i pānui 😌

Mena kei te pātai koe mō Me Pēhea te Ako i te AI , anei te tunu tino māmā rawa atu:

  • Tīmata ki ngā kaupapa matua o te ML (whakataki poto + mahi ā-Kaggle).

  • Whakamahia te scikit-learn hei ako i ngā rerengamahi me ngā inenga ML tūturu. [3]

  • Neke atu ki PyTorch mō te ako hōhonu me ngā takahuri whakangungu. [4]

  • Tāpirihia ngā pūkenga LLM mā te whakamahi i tētahi akoranga mahi me ngā tīmatanga tere API.

  • Hangaia kia 3–5 ngā kaupapa e whakaatu ana i: te whakarite raraunga, te whakatauira, te aromatawai, me tētahi takai "hua" māmā.

  • Whakaarohia te mōrearea/te whakahaere hei wāhanga o te "oti", kaua hei tāpiritanga kōwhiringa. [5]

Āe, ka ngaro pea koe i ētahi wā. He mea noa tēnā. He rite te AI ki te whakaako i tētahi tōstare ki te pānui - he mea whakamiharo ina mahi, he mea whakamataku ina kore e mahi, ā, he nui ake ngā mahi whakahou i tā te tangata e whakaae ana 😵💫

Ngā Tohutoro

[1] Ngā Tuhipoka Kauwhau a Stanford CS229. (Ngā kaupapa matua o te ML, te ako tirotiro, te anga tūponotanga).
https://cs229.stanford.edu/main_notes.pdf

[2] MIT 6.S191: Whakataki ki te Ako Hōhonu. (He tirohanga whānui mō te ako hōhonu, ngā kaupapa hou tae atu ki ngā LLM).
https://introtodeeplearning.com/

[3] scikit-learn: Aromatawai tauira me ngā inenga. (Te tika, te tika/te maumahara, ROC-AUC, me ētahi atu).
https://scikit-learn.org/stable/modules/model_evaluation.html

[4] Ngā Akoranga PyTorch – Akohia ngā Kaupapa Taketake. (Ngā Tensor, ngā huinga raraunga/ngā uta raraunga, ngā takahuri whakangungu/arotake).
https://docs.pytorch.org/tutorials/beginner/basics/intro.html

[5] Anga Whakahaere Mōrearea AI NIST (AI RMF 1.0). (He aratohu AI pono, e hangai ana ki te mōrearea).
https://nvlpubs.nist.gov/nistpubs/ai/nist.ai.100-1.pdf

Rauemi Tāpiri (ka taea te pāwhiri)

Kimihia te AI Hou i te Toa Kaiāwhina AI Whaimana

Mō Mātou

Hoki ki te rangitaki